← The index · Language, art & other · by sotoalt

TraderFly

Male CNS LIF maps sugar GRNs → BUY, looming LC4 / LPLC2 → ESCAPE sell. No PAM11 / PPL101 P&L pulse.

Repository →

MaleCNS

Fidelity · reviewed

L1 Dynamics

Neurons are simulated — leaky integrate-and-fire or better — with documented, tuned parameters.

Dispute this rating →

The evidence

No evidence, no level.

The LIF parameters are inherited verbatim from the Shiu et al. reference model and listed explicitly.
https://github.com/SotoAlt/traderfly-brain/blob/05a61a7dad295369c0ded8f030422047d8f2dca8/packages/brain/src/lif.ts#L41-L62
In the spike-propagation loop the connectome weights are only read, accumulated into a delay ring; no weight is ever assigned.
https://github.com/SotoAlt/traderfly-brain/blob/05a61a7dad295369c0ded8f030422047d8f2dca8/packages/brain/src/lif.ts#L226-L235
The decision path states that nothing is trained and nothing is tuned per token - the wiring is the connectome, read out at named motor and descending neurons.
https://github.com/SotoAlt/traderfly-brain/blob/05a61a7dad295369c0ded8f030422047d8f2dca8/packages/mechanics/src/fly.ts#L1-L20

Rating note

A leaky integrate-and-fire simulation of the MaleCNS v1.0 graph in TypeScript, with parameters taken verbatim from Shiu et al. 2024 (-52 mV rest/reset, -45 mV threshold, 20 ms membrane, 5 ms synaptic, 2.2 ms refractory, 1.8 ms delay, 0.5 ms step) and synapse weights signed by predicted neurotransmitter; market conditions become Poisson stimulation of real sugar/bitter GRNs, LC4+LPLC2 or Johnston's organ neurons, and the decision is read out of MN9, the giant fibre or aDN. Not L2: the connectome weight array is read-only in the spike-propagation loop, nothing in the repo trains anything (no optimiser, no gradient, no fitted readout anywhere in the source), and the source states outright that nothing is trained and nothing is tuned per token. Also stronger than a frozen-weights caveat - every decision starts from a fresh silent brain, so not even membrane state persists between trades. Ignore the token and the P&L framing; the simulation is what earns the level. One caveat for a reader: the exit thresholds, the trading engine and the site are deliberately not published, but the whole brain and stimulus-encoding path is.

Record

CategoryLanguage, art & other

DatasetsMaleCNS

LevelL1 — reviewed

Listed intownie/awesome-fruit-fly

First indexed2026-09-15

[livetraderfly.xyz

Nearby

All projects →

FLM

Frozen language model coupled to the retained MaleCNS graph through a trained readout adapter. Language ability comes from the pretrained language model.

L0 WIRING · MALECNS · [REVIEWED]

fly-brain

FlyWire LIF simulation with a natural-language interface and project-defined dopamine-learning experiments.

L1 DYNAMICS · FLYWIRE · [REVIEWED]

Ratings on thelearningfly.com are proposals, not verdicts. Every one of them is arguable in public.